You can always try re-flowing the solder using a hot-air rework station.

The "easy" way is to get some no cleanup liquid solder flux, get some under each chip before you re-flow it, and then re-flow it. Only do one chip at a time.

The GPU and RAM chips will be the ones you want to re-flow.

That will fix any broken solder joints which seems to be a main problem with video cards stopping working.

You can also try the bake-in-the-oven method but this usually only ends up in a temporary fix.

The more difficult way is to completely remove the chips, clean off the old solder, and then re-ball them and re-solder them to the board.

This requires the correct solder masks as well as the correct size solder balls and more experience soldering as you have to line up everything just right.

Looking at your list and thinking: is it just me or do high-end graphics card really die more often than mid-range ones?

No, in my experience, high end vs. low end isnt really an issue. However, ATI/AMD dies way more often than Nvidia hardware...at least in my personal experience.

dr_st wrote:

Looking at your list and thinking: is it just me or do high-end graphics card really die more often than mid-range ones?

In my experience, yes.
My X1950 Pro AGP died three days ago. I also have a dead Radeon 9700 (although those are known to have problems) and 6600 GT.
On the other hand, a TNT2 M64, GeForce 2 MX400 or FX 5200 will seemingly never die. I've got a lot of them from various lots and every single one works.

I think it all comes down to heat. There's a reason GPUs come with giant coolers these days.
For example, the X1950 Pro has only a slightly lower TDP than a GTX 1050, yet no GTX 1050 comes with a crappy single slot cooler. Also, VRM cooling was almost non-existent back in the day.

watson wrote:

In my experience, yes. My X1950 Pro AGP died three days ago. I also have a dead Radeon 9700 (although those are known to have pr […]
Show full quote

dr_st wrote:

Looking at your list and thinking: is it just me or do high-end graphics card really die more often than mid-range ones?

In my experience, yes.
My X1950 Pro AGP died three days ago. I also have a dead Radeon 9700 (although those are known to have problems) and 6600 GT.
On the other hand, a TNT2 M64, GeForce 2 MX400 or FX 5200 will seemingly never die. I've got a lot of them from various lots and every single one works.

I think it all comes down to heat. There's a reason GPUs come with giant coolers these days.
For example, the X1950 Pro has only a slightly lower TDP than a GTX 1050, yet no GTX 1050 comes with a crappy single slot cooler. Also, VRM cooling was almost non-existent back in the day.

And that is the reason I make sure to have very good airflow through the cases in any system I build. I haven't had a GPU die on me since the 7900GS I had back in the day that I volt-modded to be able to do a 50% overclock on. And it lasted a year before it died even though when I got it, it produced artifacts at stock speeds.

Newer systems generally have fans feeding the GPU cool air from outside the case.

What about board partners? Do you find certain ones are worse or are they about the same?

Mostly about the same. There are some specific models from different companies that had/have issues.. mostly due to poor cooling and/or poorly designed/inadequate power circuitry.

dr_st wrote:

Looking at your list and thinking: is it just me or do high-end graphics card really die more often than mid-range ones?

GPUs in general run at significantly higher temperatures than CPUs and many other components do. Naturally the higher-end GPUs are going to be more complex internally, pull more current and run even warmer, and yet there is only so much copper/aluminium that one can conceivably cram into the space of 1~3 expansion slots; CPU HSFs on the other hand tend to not be so hamstrung for space. Then, the newer lead-free solder is not as "elastic" as the old solder and, on a GPU, it is nonetheless subject to a significantly wider gradient of temperatures than a typical other BGA-mounted component (such as a northbridge chip) would be; finally, the sagging and warping of a PCB from the weight of the necessary gigantic heatsink adds insult to injury. With all of this in mind, it's not surprising (at least to me) that high-end GPUs tend to fail noticeably more often than many other PC components.